An Investigation of the Mood in the Media
Where We Started
People’s moods differ throughout time. Sometimes, we are affected by personal matters, social affairs, or the weather. Other times, more prominent aspects might affect our mood; we can be affected by the outcome of political elections, international disputes, or long-lasting pandemics. However, we constantly talk about our mood and its causes.
Provided with the QuoteBank data set, we became interested in what it could reveal about the truth in the everyday talk about our mood. For example, does the temper become better towards the weekend, or do we all love Mondays deep inside? As the weather probably affects our mood, is the mood better during the summer and worse when the cold hits us in the winter? Moreover, as the data set consists of quotes from the media, what can we say about the trends of the mood in the media throughout time?
Based on our initial questions, our thoughts wandered into what we could say specifically about the mood in the media. As the media sets the agenda for the public debate, it is interesting to see how much positivity and negativity reach the readers’ minds. Which media outlets provide us with the most positivity, and which are more negative? Moreover, can we see any differences across the subsets of speakers in the data set? For example, are politicians in position more positive in their quotes, and are those in opposition more negative? Do women tend to be more optimistic than their peers?
We aim to present our findings on the abovementioned topics in the upcoming data story. Starting with the QuoteBank data set provided by dlab @ EPFL, we will utilize information from Wikidata to research our questions by sentiment analyses.
Sentiments Analyses – What Are They?
Sentiment analyses aim to identify and extract subjective information in text, telling whether the phrase is positive, negative, or neutral. Using libraries, one could input a sentence, such as a quote from the media, and get a score. This will be our primary tool in finding trends in the mood of the media.

Compound Scores – Huh?
In our analyses, we have utilized the compound score of the quotes in Quotebank. This is a measure of the total sentiment of a quote, indicating the sum of positive, negative and neutral scores of the words in a text, eventually normalized to be a value between $-1$ and $1$. A score of $-1$ indicates the most negative compound, while a score of $+1$ indicates the most positive.
Fear itself is always more dangerous than the thing you fear. The fear of death is worse than dying. Fear takes you hostage and kills your resistance. Nowhere is fear more fatal than in prison. Mehmet Altan
The above quotation has a compound score of $-0.9891$, indicating that it is quite negative, as you probably could predict when reading it.
I was glad to see him, and I'd like to think he was glad to see me. L. Kennedy
On the other hand, the above quotation receives a compound score of $0.8176$, which is reasonable as it contains some light, positive words. While these quotations show the extremes on the compound scale, the majority of quotations lie in the middle. For instance, the following quotation gets a score of $0.00$, and is therefore considered absoluteley neutral:
There's a fact of life. The Assad regime has stabilized the situation in much of Syria. Eran Lerman
As you might have thought of, the art of dedicating compound scores to quotations is not perfect. The machine learning methods generating such scores give different aspects of the given text a score and averaging it over the sentence. Such an algorithm could be sensitive for some extremely positive or negative words, while other while be given a lower weight. On the other hand, the algorithm is likely not to discover the underlying context of a quotation, and could thus mark a quotation of positive, even if it is considering a deeply serious matter.
Subjectivity Scores
- This is Our Data – Milestone 2
- Does the Mood in the Media Differ Throughout Time?
- Does the Mood in the Media Differ Across Media Outlets?
- Do Men and Women Differ in Mood?
- Are Politicans More Negative or Positive Than Their Peers?